Impact of Hurricane Katrina on Roadways in the New Orleans Area Technical Assistance Report No. 07-2TA by Kevin Gaspard, Mark Martinez, Zhongjie Zhang, Zhong Wu LTRC Pavement Research Group

Conducted for Louisiana Department of Transportation and Development Louisiana Transportation Research Center

The contents of this report reflect the views of the authors who are responsible for the facts and the accuracy of the data presented herein. The contents do not necessarily reflect the views or policies of the Louisiana Department of Transportation and Development or the Louisiana Transportation Research center. This report does not constitute a standard, specification or regulation.

March 2007

ii

ABSTRACT On August 29, 2005, Hurricane Katrina devastated New Orleans and southeastern Louisiana, leaving hundreds of thousands either displaced or homeless. Nearly four weeks later, Hurricane Rita made landfall in the southwestern portion of the state, further damaging Louisiana’s infrastructure and impacting the New Orleans area. In response, LTRC personnel conducted pavement testing on several on-going construction projects that were submerged to determine if contract modifications would be necessary to address damage impact. Damage was found in asphalt and concrete layers, and subgrades were found to be very weak. For one project, LA 46, LTRC had “before and after” data which indicated that the damage incurred was equivalent to three inches of asphalt concrete. As a result, LaDOTD contracted with Fugro Consultants, LP, to conduct testing on 238 miles of state highways in New Orleans at 0.1 mile intervals. Fugro conducted Falling Weight Deflectometer, Ground Penetrating Radar, and Dynamic Cone Penetrometer testing along with coring selected locations for thickness and damage verification to determine the extent of structural damage to these pavements. Because there was no “before” data, a traditional forensic type analysis could not be undertaken. With the use of GIS mapping and NOAA flood mapping, data points could be identified as either submerged or non-submerged. The non-submerged data were then considered as a control set, and the submerged data were considered as the experimental set. In this manner, the data could be tested using standard analysis of variance techniques to test the hypothesis that the submerged pavements were weaker and therefore damaged as a result of the hurricanes. It is noted that this methodology does not imply that the nonsubmerged pavements were not damaged also, but provides a relative damage estimate. Once weaker strength parameters were determined, standard pavement design methods were applied to the structural numbers and subgrade modulii to determine an equivalent amount of asphalt concrete for this strength loss.

iii

In general, it was found that asphalt pavements had strength loss equivalent to about two inches of new asphalt concrete and that thinner asphalt pavements were weaker than the thicker pavements. Very little relative damage was detected for the PCC pavements. The composite pavements demonstrated no need for additional structure in the pavement layers; however a weaker subgrade for the submerged areas equivalent to nearly one inch of asphalt concrete was identified. Using recent bid prices in New Orleans of $250,000 per mile for a typical rehabilitation scenario (mill four inches/replace four inches of asphalt concrete), an estimated cost for the approximately 200 miles of submerged state highway pavements would be $50 million. There are another 300 miles of federal-aid and 1500 miles of non-federal aid roads that were submerged in the New Orleans area.

iv

TABLE OF CONTENTS Abstract .............................................................................................................................. iii Table of Contents.................................................................................................................v Introduction..........................................................................................................................1 Methodology ........................................................................................................................3 Discussion of Results.........................................................................................................15 Conclusions........................................................................................................................25 Appendix 1.........................................................................................................................27 Appendix 2.........................................................................................................................47 Appendix 3.........................................................................................................................61 Appendix 4.........................................................................................................................65

v

vi

INTRODUCTION On August 29, 2005, Hurricane Katrina devastated New Orleans and southeastern Louisiana, leaving hundreds of thousands of people either displaced or homeless. Nearly four weeks later, Hurricane Rita made landfall in the southwestern portion of the state, further damaging Louisiana’s infrastructure and, once again, bringing destruction to portions of the New Orleans area. While much of the damage to buildings and bridges was immediately obvious, the damage imparted to roadways would not be so easy to recognize. It was expected that the sustained flooding (three days or more) had damaged the roadway pavement structures below the surface in the submerged areas. Further, subsequent debris removal was expected to provide additional damage both to roads that were submerged and those that were not submerged. Such damage, if not repaired, is certain to have a profound impact on the recovery and future social and economic development of the New Orleans area. LaDOTD’s data collection efforts prior to the storm were designed largely to address pavement management and rehabilitation efforts and were, therefore, not suitable for evaluating the pavement structural damage that resulted from the submergence. Pavement distress data such as International Roughness Index (IRI), rutting, and cracking were available on some state routes, but vital structural data needed to determine the flood’s impact, such as the resilient modulus of the pavement layers or overall structural number (SN), were not collected as part of the pavement management systems’ biennial program. Such limitations meant that it would be impossible to conduct a comprehensive “before and after” style structural analysis to determine the reduction in strength of the pavement layers caused by the flood. The Louisiana Transportation Research Center (LTRC) initially conducted structural damage testing on several roads that were under construction to determine any damage that might require additional work. Based on the preliminary results additional roads were tested in the New Orleans area. In all, a total of eight roadways were tested consisting of Falling Weight Deflectometer (FWD), Dynaflect, Dynamic Cone Penetrometer (DCP) and coring. The FWD provided pavement modulus and subgrade modulus through a back calculation method; the Dynaflect provided pavement structural number (SN) along with subgrade modulus; the DCP provided verification of the base and subgrade readings; and, the coring provided thicknesses

1

and verification of moisture damage. Appendix A presents test results and a summary analysis for each project. As presented in Appendix A, structural damage is indicated in the asphalt concrete pavement, concrete pavement, and base course along with weakened subgrades in most cases as related to the modulus of these layers. Some of the concrete pavements are showing voids at the joints with lost joint transfer efficiencies. In seven of eight cases there is no “before hurricane” data such that actual damage can be directly attributed to the submergence. However, there is “before” data for one project, LA 46, coming from a research project. In addition, the structural numbers (SN) for pavements under construction at the time of Katrina (I-610 and Metairie Road) are lower than the design SN, indicating damage directly caused from the submergence. A direct “before-after” analysis of structural strengths for LA 46 is included in Appendix A. On this four-lane section (CSLM 1.4 to 3.1), testing provides an average reduction of 0.9 SN and a reduction of 1.6 ksi for the subgrade resilient modulus. The results of this testing indicated that the pavement structure had been adversely impacted by the flood waters equivalent to three inches of asphalt concrete. On the basis of the LTRC investigation, LaDOTD contracted with Fugro Consultants, LP to conduct structural testing on 238 miles of state highways in New Orleans. FWD testing was undertaken every 0.1 mile and was correlated with Ground Penetrating Radar (GPR) test data which was collected continuously (0.5 foot intervals). In addition, DCP and cores were periodically taken to verify thicknesses and base course/subgrade moduli. This report evaluates the data obtained from the Fugro testing.

2

METHODOLOGY

FUGRO DATA Fugro consultants performed tests on 238 miles of federally supported urban system pavements in New Orleans. Non-destructive testing was performed with the FWD every one-tenth of a mile. GPR was used to determine pavement layer thicknesses and identify areas with thickness variations or potential voids. GPR thickness data were calibrated by conducting coring tests through the pavement and base course with thickness measurements being taken of each layer. Coring also provided information to determine the type of pavement layers (i.e., asphalt, concrete, brick) as well as their condition, and the type of base course (i.e., soil cement, sand, sand shell). A visual survey of the subgrade was conducted to determine its type (i.e, sand, clay). DCP tests were conducted to provide an additional assessment of the base course and subgrade as well as to validate subgrade strength readings from the FWD. The Fugro report provided the following data to LaDOTD. Appendix B provides details of the test factorial, descriptions of equipment used, analysis equations, and procedures. ¾ Subgrade resilient modulus (Mr) ¾ Effective pavement modulus (Ep) ¾ Modulus of subgrade reaction (k) for concrete pavements ¾ Effective structural number (SNeff) based on deflections for flexible pavements ¾ Surface curvature index (SCI) values based on surface deflections ¾ California bearing ratio (CBR) values from DCP tests results ¾ Deflection basin analysis ¾ Dynamic cone penetrometer index (DCPI) The FWD data provided by Fugro was reviewed to check for calculation and equipment errors. In order for the collected data to be considered valid for FWD testing, the deflections measured by the sensors were required to decrease as the distance of the sensor from the load plate increased. Any points collected that indicated “non-decreasing deflections” were considered invalid and removed from the data set.

3

It was decided to use the FWD data for the analysis conducted for this report in order to provide a timely assessment of the structural damage of the submerged pavements. Specifically, three test parameters, the first sensor deflection (D1), effective structural number (SNeff), and subgrade resilient modulus (Mr) were selected. ¾ First sensor deflection (D1): The deflection of the pavement at the load plate reflects the strength of the overall pavement structure. High deflections represent weaker pavement structures. ¾ Effective structural number (SNeff): The effective structural number represents the effective structural strength of the existing pavement and base course, which in this case was derived using formulas from the 1993 AASHTO design guide, deflections obtained from FWD testing, and pavement layer thicknesses determined by the GPR and validated through coring. ¾ Subgrade resilient modulus (Mr): The resilient modulus was derived using the AASHTO formula and deflection data from the FWD. The Mr was reduced by (0.33) to correlate to laboratory derived Mr as suggested in the AASHTO design guide. RESEARCH APPROACH Because there were no direct “before and after” comparison sites available other than LA 46 to show damage caused by Katrina, another methodology was chosen to demonstrate that structural damage had been incurred by those pavements subjected to submergence. The Fugro data set was divided and coded to distinguish those pavements that were submerged and those that were not. The non-submerged pavements could then be treated as a control section to test the hypothesis that damage was done to the submerged pavements. It should be noted that the control section designation does not imply that the non-submerged sections were not also damaged. In addition, further damage because of the volume and loads of the debris haul trucks that continue to travel over these weakened structures can not be determined from this data set as the Fugro evaluation was conducted several months after the waters had receded. A subsequent sampling of the same pavements in the future might reveal this additional damage. This investigation primarily attempts to determine if submergence increased the distress in the flooded areas to a greater degree than that in the non-flooded areas.

4

Methods employed to separate flooded areas from non-flooded areas ArcGIS, a commercially available global information system (GIS) software package was used to import GIS referenced maps, data points, and perform basic spatial analysis. Suitable maps for the New Orleans area were downloaded into ArcGIS from the United States Geological Survey (USGS) (http://seamless.usgs.gov/website/Seamless/). All test points from the Fugro data set were tagged with their respective GPS coordinates at the time of testing and were also imported into ArcGIS. Figure 1 shows the results of this integration. The red circles identify Portland Cement Concrete (PCC) pavements, the yellow circles identify AC pavements, and the blue circles identify composite pavements. Detailed flood maps from FEMA (http:www.gismaps.fema.gov/2005pages) were imported into the ArcGIS system to separate the flooded areas from the non-flooded areas. As an example, Figure 2 presents the FEMA map which represents the maximum extent of flooding on September 2, 2005, and Figure 3 presents the point segregation. Determining flood durations: The National Oceanographic and Atmospheric Association (NOAA) produced a series of modified False Color Infrared SPOT images (http://www.nhc.noaa.gov/) that were used to determine the duration of flooding. The specific images utilized from NOAA included images from August 31, and September 3, 5, 8, 10, 12, 14 to 20. The ArcGIS renderings for the dates of September 3 and 14 are provided in Figures 4 and 5 to serve as examples and to illustrate the ponding/differential de-watering effect described. For example, comparing Figures 4 and 5 clearly shows that the pocket of water south of Lakefront Airport in east New Orleans was drained much faster than the pocket south of Lake Pontchartrain in the Gentilly area. Using these maps, it was possible to segregate the test data according to flood duration. The datasets were segregated into four flood duration groups. The first group represented points that had been submerged for a period of one week. The second group included points that were submerged for two weeks. The third group remained under water for a period of three weeks. The final group did not flood at all. One difficulty that the researchers encountered related to image drop-outs over time. Loss of coverage in some cases made it impossible to determine when certain test points became dry. Re-examination of Figure 4, for example, does show proper coverage of St. Bernard Parish on September 3, but the detail 5

provided in Figure 6 shows that NOAA was no longer monitoring St. Bernard Parish by September 14. Data from pavements that dropped out of coverage were removed as the duration under water could not be established. Although such issues were problematic, there was enough coverage over time and over a large enough area to perform a proper analysis. ANALYSIS Statistical Methods Statistical Analysis Software (SAS) version 9.1.3 was used for hypothesis testing. Analysis of Variance (ANOVA), and a comparisonwise test of the means were used for evaluation. For this study, three structural parameters from the Fugro data were analyzed: D1, SNeff, and Mr. A confidence level of 95 percent was used for all testing. The initial data set, identified as flooded and non-flooded (flood type), included all pavement types (PCC, Asphalt, Composite). The data were then further broken down into additional factors including pavement thickness and duration of submergence. The AC pavements were divided into < 7 in., 7 to 11 in., >11 in. thick groups’, composite pavements were divided into < 16 in., > 16 in. thick groups’, and PCC pavements were divided into < 10.5 in. and > 10.5 in. thick groups. The duration of submergence was identified as non-flooded, one week, two weeks, and three weeks or more. Typically the data set for duration of submergence was smaller because of coverage drop out. An Analysis of Variance (ANOVA) test was conducted using the main factors pavement type, flood type and their interactions. Further analyses were conducted on thickness and duration of flooding. It was hypothesized that thinner pavements or pavements that were under water for longer periods of time would be more damaged than thicker sections or shorter duration. This analysis was conducted for each pavement type.

6

Figure 1: Integration of Fugro points with geo-referenced backplane in ArcGIS

7

Figure 2: September 2nd flooding

8

Figure 3: Segregation of Fugro points according to September 2nd flooding

9

Figure 4: September 3 flooding (NOAA)

Figure 5: September 14 flooding (NOAA) 10

Figure 6: Detail of St. Bernard Parish on September 14 (taken from Figure 5)

11

Pavement Damage Analysis LaDOTD has historically used Dynaflect-generated SN values for pavement design. AASHTO pavement design coefficients were developed based on Marshall Mix design properties as correlated to SN. FWD derived SN has not been used in Louisiana. In order to convert the SN derived from the FWD to Dynaflect SN, a correlation equation was developed using data from previous projects. Three hundred and thirty-four (334) points from 15 projects were used for this correlation. Appendix C provides the development of the correlation. If the statistical analysis demonstrates that the submerged pavements by pavement type had structural damage through lower SN, then this loss of SN can be converted to an equivalent pavement thickness, using the coefficient for asphalt concrete to represent the structural loss. For example, if flooded pavements had an SN of 3 and the non-flooded pavements had an SN of 4, then the loss of structure is an SN of 1. In Louisiana, a layer coefficient of 0.44 is generally assigned to asphalt mixtures. Therefore, the equivalent in. of AC due to the distresses in this example would be 2.3 in. (1/0.44). This conversion of SN to equivalent inches of asphalt concrete (AC) provided a mechanism to estimate a cost associated with the damaged pavements. In addition, to weakened pavement structures, weakened subgrades should also be considered. The DARWIN pavement analysis procedure is used to infer the damage caused by decreasing the subgrade resilient modulus. DARWIN 3.1 is the designation for a series of AASHTO's computer software programs for pavement design and was part of the implementation of the 1993 AASHTO Guide for Design of Pavement Structures. DARWIN 3.1 is divided into four modules: Flexible Structural Design, Rigid Structural Design, Overlay Design and Life Cycle Cost. Each module addresses a specific item in the overall pavement design process. The thickness of pavement structures for design purposes is typically determined using the DARWIN computer program. The main input parameters for the program are the highway’s subgrade resilient modulus (Mr) and the traffic loads. The output from the analysis is the required structural number (SN) that the design pavement would need in order to support the intended traffic and protect the subgrade. 12

Using the output from the DARWIN pavement design, a curve can be created that shows how SN would vary as the subgrade Mr is changed for a given traffic load level. Figure 7 provides an example using data generated from the LA 46 pavement which is the only pavement in the study for which before and after data exists (example DARWIN tables are presented in Appendix D). If the subgrade resilient modulus is 6 ksi, then the required pavement thickness would be 10.5 in. If the subgrade resilient modulus is reduced to 3 ksi, then the required pavement thickness would be 13 in. Using this methodology, an equivalent amount of pavement thickness can be assigned due to the reduction of subgrade modulus. In this example, it would be 2.5 inches of pavement.

y = 18.476x -0.321 R2= 0.9983

LA 46 Subgrade Mr AC thickness (in) 18.0 17.0 16.0 15.0 14.0 13.0 12.0 11.0 10.0 9.0 8.0 0.0

1.0

2.0

3.0

4.0

5.0

6.0

7.0

8.0

9.0

10.0

11.0

12.0

Subgrade Mr (ksi) Figure 7 AC verses subgrade Mr (LA 46)

It is recognized that this methodology for the damage attributable to lower subgrade modulus is project-specific because of the dependence on traffic loading. However, if extended to all the pavements submerged, it can provide at least an estimate of the total damage. Once a particular project has been identified for rehabilitation, the actual traffic can be used to determine the thickness required.

13

Figure 8 presents a flow chart of the statistical methods and methodology used to analyze the Fugro data.

Flooded AC

PCC Non-flooded

Composite

Remove erroneous data

Flooded AC Composite

All points

PCC Non-flooded Filtered points

Perform ANOVA analysis on pavement types using three parameters (D1, SNeff, and Mr ) to test for differences. If found, test each pavement type separately. AC

PCC

Composite

Run an ANOVA test on each pavement type for three groupings: flooded vs. non-flooded, pavement thickness, flood duration. AC

PCC

If differences are found, then perform an analysis to translate distress into equivalent thickness of AC.

Figure 8 Analysis flow chart

14

Composite

DISCUSSION OF RESULTS

STATISTICAL ANALYSIS Analysis of Pavement Type and Flood Type A two-way ANOVA was performed on the entire data set for each of the test parameters. The main factors were pavement type and flood type with the interaction also examined. Table 1 presents the p-values. P-values less than 0.05 indicate that the factors are significantly different. These values have been bolded in the table. All parameters are significant for the two main factors of pavement type and flood type. This indicates that for each parameter, the pavement type and whether or not the pavement was submerged, the values do not come from the same data set; that is, they are different. The interaction factor shows no significant difference for the Mr parameter. This makes sense as the Mr parameter is probably independent of the pavement type. Because of this influence of pavement type, the remaining analyses were conducted by pavement type. Table 1. Two way ANOVA for Pavement Type and Flood Type p-value (n=2274) Pavement

Flood

Ptype

type

type

*Ftype

D1

10.5" Thickness Group

PCC

< 10.5" > 10.5" Thickness Group

PCC

< 10.5" > 10.5"

n 265 176

Flooded mean std dev 5.17 1.90 4.78 2.07

D1 Non-flooded n mean std dev 4 4.24 0.76 41 4.53 1.09

n 265 176

Flooded mean std dev 7.67 1.57 8.30 1.52

SNeff Non-flooded n mean std dev 4 8.27 1.27 41 8.40 1.19

n 265 174

Flooded mean std dev 5.33 1.55 5.85 1.89

Mr Non-flooded n mean std dev 4 6.12 0.72 41 5.55 1.15

p-value 0.3322 0.4842

p-value 0.4496 0.8851

p-value 0.3113 0.3714

Table 7 Means Test for PCC Pavement Thickness Level 1 2

Thickness Range 10.5” (n=217) Note

D1

SNeff

Mr

A B “A” means highest D1

B A “A” means highest SN

B A “A” means highest Mr

Analysis of Composite Pavements Three factors were analyzed for the composite pavements including flood type, thickness, and duration of flooding. Table 8 presents the results from the ANOVA tests. These results are more complex than either the asphalt or PCC pavements. Table 9 presents the 19

ANOVA by pavement thickness. Table 10 provides the overall structures that represent these pavements. There is much masking of results occurring because of the relative strengths of the asphalt, PCC, brick and base course materials that constitute these pavements and their overall thicknesses. For the first time, duration of submergence provides significant differences in performance of D1 and Mr. These same parameters also show performance differences between flooded and non-flooded pavements. The SNeff parameter seems to be masked by the strengths of the PCC layers although the thinner sections demonstrate significant difference in performance.

Table 8 Composite Pavement ANOVA Testing p-value Thickness (n=907) 0.2597 0.0084 0.1101

Flood type (n=907) 0.0002 0.5125